One cumulative computing and AI pathway. Students begin with precise logic, prediction, testing and debugging, then grow into text programming, data, machine learning, software architecture and independent intelligent-system engineering.
8 Bands • 12 Projects per standard Band • Increasing learner ownership every year
The point is not to make the tools harder every year. The point is to make the learner more capable of deciding what to do, why to do it, how to prove it works, and eventually whether it should be built at all.
A young learner does not begin with "AI tools." They begin with the deeper idea beneath every piece of computing:
I have an intention. I need to represent it precisely enough that a machine can act on it.
By the end of Mera, learners are working with state, rules, data, testing and simple learned models—but the visual medium keeps syntax from stealing attention from the thinking.
Lobuche makes the representation more demanding without changing the underlying discipline. The learner begins to see that the same algorithm can exist as a diagram, pseudocode, blocks or text. By Lobuche Summit, text code is no longer a foreign language. It becomes one of the normal ways the learner represents and tests an idea.
Ama Dablam shifts the learner from solving bounded computational tasks toward building purposeful systems. Data becomes real material. Algorithms become choices. Models are trained and evaluated. Programs become modular. By Class 9, the learner increasingly decides how the system itself should be structured.
Pumori gives the learner a serious defined problem—but not the architecture. The student must choose data structures, services, databases, APIs, pipelines, models, LLM components and bounded agents where appropriate, then test the system under failure and defend the engineering decisions.
Cho Oyu adds the final responsibility shift: the problem is no longer neatly packaged. The learner investigates reality, identifies stakeholders, decides what should and should not be automated, defines success and unacceptable failure, chooses the technical approach, builds the system and defends the entire decision chain.
It is not eight disconnected coding courses. Each Band assumes, retrieves and extends what came before. The tools mature, the systems become more sophisticated, and—most importantly—the learner owns more of the thinking every year.
Make the machine do exactly what you meant.
A student new to AI Pathway always enters at their stage's year-one Band. The year-two Band is for students who completed year one.
A student who joins in the second Class of a stage does the year-one Band that year, then rejoins their Class group at the next stage. They do not repeat a year, and they do not skip ahead into a Band that assumes work they have not done.
Mera Peak is the only Band using Peak. Class 11 is Pumori Summit. Class 10 is a deliberate pause, not a missing year — reactivation is not reteaching the pathway from zero.
A Class 4 learner should not be asked to invent a software architecture. A Class 12 learner should not be given one. AI Pathway deliberately transfers responsibility as capability grows.
The starting point. Move along the Bands to watch responsibility transfer.
Supplied
Mostly supplied
Strongly guided
Important bounded choices
Constructs within known visual tools
Learns to predict, test and compare expected vs actual
Guided method becomes a habit
Not yet a learner-owned target
Begins with clarity, fairness and simple user impact
A later Band is not just a harder tool. It is a larger share of the problem becoming the learner's responsibility.
The same habits return at greater scale, lower support and higher consequence. Debugging does not disappear when AI arrives. It becomes more important.
The same question, asked of a system that keeps getting larger. Step through the years.
Why did the robot turn the wrong way?
The habitCompare what I expected with what actually happened.
Eight short interactions, one per Band. Each takes under a minute and shows the kind of reasoning the year develops — not a lesson from the curriculum.
A computer cannot fill in the missing steps the way a person often does. Build a precise route, predict what happens, then run it.
Column 1, row 3 · facing North
A Band is not a pile of unrelated activities. Each standard Band contains 12 Projects arranged so that capability is established, combined, transferred and increasingly owned.
Build the representations, habits and new capabilities the Band needs.
SynthesisP4 combines the important work from P1–P3.
Coordinate established capabilities in larger systems, less familiar combinations and lower support.
SynthesisP8 proves a materially stronger middle-Band integration.
Increase novelty, transfer distance, strategy ownership and technical defense.
SynthesisP12 is the Band capstone—the end-state, not merely the flashiest build.
A standard Project contains 18 Checkpoints. A standard Checkpoint contains 10 Activity Nodes. That structure exists to create meaningful learning transitions—not to give parents 180 boxes to scroll through.
What this page shows by default
Band → Project → developmental transition → proof
Deeper signed-in exploration may reveal selected Checkpoint transitions. Do not expose all Activity Nodes by default.
A student can produce a beautiful result with heavy support, copying or AI assistance. AI Pathway therefore separates:
Parent proof should compress rich internal evidence into a few understandable facts about what the learner actually did.
A visual sequential program controlling Koda through the final mission context.
This is what a parent normally sees at the end of a project — and it is the part that proves the least.
This viewer demonstrates the type of evidence AI Pathway preserves. Do not present sample evidence as belonging to a real learner unless a real approved evidence packet is supplied.
Class 10 is protected because of examination pressure. AI Pathway does not force a normal 12-Project Band into that year, and it does not restart Class 11 from zero.
Ama Dablam Summit evidence and artifacts remain part of the learner history.
No standard Band is assumed.
Pumori Summit retrieves → diagnoses → repairs gaps → re-establishes fluency → accelerates.
Reactivation is not reteaching the pathway from zero.
See how Pumori restarts the climbCompare any two Bands on the responsibilities the curriculum actually transfers. It opens on the first year of the climb against the last.
Yes—progressively and intentionally. Classes 4–5 use visual or equivalently low-friction programming so syntax does not crowd out computational reasoning. Class 6 deliberately bridges to text, Class 7 makes text primary, and later Bands use increasingly credible software-engineering environments.
No. AI Pathway develops algorithms, programming, data, systems, debugging, testing, software engineering and machine learning. Generative AI and agents appear inside that broader technical foundation, and students are expected to test, constrain and sometimes reject AI-generated output.
Because the early target is computational thinking, not punctuation. A visible, high-feedback representation lets younger learners reason seriously about algorithms, state, decisions, debugging, testing, data and model behavior. The same ideas are then deliberately translated into text.
AI Pathway separates Project completion from capability mastery. Important evidence can include prediction, tracing, debugging, tests, transfer to a changed context, explanation/defense and the exact support used. AI-assisted work is credited only for capabilities the learner actually demonstrates.
It gets more sophisticated, but the deeper progression is ownership. Younger learners receive the problem and structure. Later learners increasingly choose the algorithm, architecture and evaluation strategy. By Cho Oyu, they increasingly own the problem itself.
There is intentionally no standard Band because of examination pressure. Prior evidence remains part of the learner history. Pumori Summit in Class 11 begins with targeted retrieval and reactivation, then accelerates into advanced engineering.
AI assistance can become normal, but accountability does not disappear. Students must be able to understand important logic, test it, trace dependencies, modify it, find flaws, reject inappropriate output and defend the final system.
No. AI-vs-non-AI judgment is itself a capability. A mature learner may correctly choose a deterministic workflow because it is simpler, safer, cheaper or more reliable.
The pathway is informed by contemporary computer science education references, including the 2026 CSTA foundational progressions, while following SCC's own cumulative capability architecture. CSTA is a reference point, not an endorsement or the organizing structure of the product.
Informed by contemporary computer science education standards, including CSTA 2026, and extended through SCC's cumulative AI Pathway capability architecture.
The 2026 CSTA foundational progressions reinforce several important directions already built into AI Pathway: algorithms before code, program comprehension, variables/state, debugging/testing, data, machine learning, evaluation of AI-generated output and human impacts. AI Pathway uses CSTA as one external reference while preserving its own cumulative Band progression and responsibility staircase.
We'll centre the pathway on what they are ready to learn now, what prepares them, and what comes next. Two Classes share every stage, so we also ask whether this would be their first year.
The destination is a learner who can understand a problem, represent it, build a system, test the evidence, diagnose failure, make technical judgments—and eventually decide what should be built in the first place.
The pathway runs at Chelsea World School, 2 afternoons a week.